Decision records
Automated Decisions and Constitutional Accountability
A decision does not become less public because software helped produce it. If the outcome affects recognition, service, liberty, dignity or remedy, the record should show who controlled the rule, what input was used, what output was produced and which human authority can review it.
Where failure occurs
The archive repeatedly distinguishes disposal from resolution. In automated environments, the same problem can appear as a closed ticket, rejected application, failed match or generated risk label without a meaningful explanation to the citizen.
Minimum record
- the rule or model used for the decision;
- the public authority responsible for the system;
- the source record or input that affected the citizen;
- the human review route and correction process;
- the preservation of logs needed to test the decision later.
What remains unresolved
This page does not assert that a named automated system is unlawful. It sets the evidentiary standard the archive will use when a future case file concerns automated public decision-making.
Why automation changes the accountability question
Traditional administrative decisions are made by identifiable officials whose authority flows from a statute, order or policy instrument. When an automated system replaces that official — or mediates the decision by ranking, scoring, matching or flagging a citizen — the accountability chain can become invisible. The output may be indistinguishable from a human administrative order, but the reasons for it may be locked inside a software model that no public officer can explain in terms a citizen can challenge. Article 12 of the Indian Constitution defines the State as encompassing the Government of India, state governments, Parliament, state legislatures, and other authorities within India's territory or under its control. The DISHA research position is that an automated system operating under the direction or control of any such authority does not become exempt from constitutional accountability simply because its logic is computational rather than human.
Disposal versus resolution
The DISHA archive repeatedly distinguishes between disposal and resolution. A complaint or application is disposed when a system generates a closure status: ticket closed, application rejected, match not found, grievance resolved. It is resolved only when the underlying question has been genuinely addressed and the citizen has received an outcome that corresponds to their rights and record. Automated environments accelerate disposal. They do not, by themselves, produce resolution. A citizen who receives a risk label, a benefit denial, a service refusal or a failed match from an automated system has received a disposal. Whether that disposal is constitutionally adequate depends on whether an explanation was provided, whether the source record was accurate, whether a human authority reviewed the case, and whether a correction and remedy path is accessible.
The evidence standard for automated-decision cases
When the DISHA archive examines a case involving an automated public decision, it asks for the minimum record described above: the rule or model, the responsible authority, the source input, the review route and the evidence preservation log. In addition, the archive will ask whether the citizen was informed that an automated process was involved. Undisclosed automation in public decision-making raises its own constitutional concern: a citizen cannot challenge a reason they have not been told exists. The archive will record the absence of disclosure as an unresolved question under the DISHA claim-to-source system, pending a public authority response.
Connection to Digital Constitutional Personhood
The framework of Digital Constitutional Personhood, developed as part of this research, holds that a citizen's digital identity is an extension of their constitutional personhood under Articles 14, 19 and 21. When an automated decision misidentifies, mislabels or denies a citizen through their digital representation, the harm is not merely administrative: it is a constitutional event. The right to equality under Article 14 is engaged when automated systems apply opaque criteria that cannot be tested for consistency. The right to dignity under Article 21 is engaged when a citizen is reduced to a data profile that produces a denial without a face, a name or an explanation. The accountability standards for automated decisions are therefore inseparable from the broader constitutional accountability framework that Article 12 anchors.